Apogee AD-1000 Operating Manual page 47

AD-1000 Operating Manual
Something For Nothing?
If we take a well-recorded 16-bit digital audio source and we decide to add some digital EQ, compression, gain
change, de-click, de-crackle or even the latest surround-sound process, we are digitally manipulating our 16-bit
numbers. We don't get more audio information out than we put in, but we do get numbers with resolution
greater than the 16 bit input signal. These extra 'detail' bits we pick up contain some of the results of whatev-
er process we performed to our original 16-bit audio source – and ideally we should hang on to those bits. They
show up as improved smoothness, detail, image and depth. The aim of the various encoding schemes is to hold
on to the extra resolution after digital processing or A to D conversion when transferring to a 16-bit CD quali-
ty output.
Holding on to more bits in a 16-bit CD world
An ideal 'Super CD' system would take as much as 24-bit resolution digital audio and capture the same detail
and quality on to our 16-bit CD. We don't live in an ideal world, but it is possible to capture much of the added
detail in 20-bit (and greater) systems and bring it into the world of 16-bit DATs and CDs.
Dancing Bits On The Noise Floor
All the encoding systems make the last digital bits dance so they capture extended resolution in the 16-bit CD
format. A useful way to separate the different processes (dance steps?) is to look at how each handles the noise
floor:
• Common dither methods compromise the 16-bit noise floor – they add noise
• Noise shaping and 'bit mapping' trade a reduced noise floor for a large boost at high frequencies
• UV22, Apogee's proprietary process, keeps the audible noise floor solid at the theoretical minimum for 16-
bit systems
Although noise shaping and bit-mapping systems (questionably) focus on the noise floor, users often hear this
'improved' noise floor as changing with the music, making it watery and 'fluid-like'. Traditional dither adds noise
and raises the noise floor. UV22, on the other hand, presents a constant, smooth and stable noise floor, unob-
trusively at the theoretical minimum level, but through which can be heard full 20-bit detail.
How Does It Compare To Analog?
We all know one of the main reasons for going digital: low noise. So why do some engineers still master to ⁄
1
2
inch analog? Technically, analog recordings may appear to be limited by their noise 'floor'. On closer listening,
however, the noise floor turns out not to be as solid as the name suggests. A better analogy would be to com-
pare the 'floor' to the surface of a crystal-clear lake, where you can see right into the depths. Analog noise is
like that: smooth and constant – but you can hear through it.
This is where digital has differed in the past. The (albeit low) noise floor truly was a limit – more like a stirred,
muddy lake. Dither has been used (intentionally and unintentionally) for years to clarify the grunge usually lying
on the bottom. The problem, however, was that dither is quite inefficient at capturing full fidelity musical detail,
because it is very slow at its job – and it invariably increases the noise.
UV22 is the most efficient method of all in capturing extended resolution into the 16-bit format. This powerful
information carrier sits inaudibly out of hearing, yet presents a smooth, white, unvarying noise floor through
which can be heard undistorted detail up to 30dB lower in level – extending full- fidelity information beyond 20-
bit resolution to your 16-bit CD.
Impressive comments from critical listeners
"The reverb detail and stereo spread are amazing."
—Michael Bishop, Telarc International, Cleveland
Engineers Michael Bishop, Scott Burgess and Elaine Martone tested a number of systems: Apogee UV22
Super CD Encoding; Sony Super Bit Mapping; Gambit; Harmonia Mundi; and Sonic Solutions Turbo Bit
Mapping. They used a recording of the Atlanta Symphony Orchestra and Chorus, conducted by Yoel Levi, per-
forming Ravel's Daphnis and Chloe (November 1993 release, Telarc catalogue number CD-80352). The source
was 20-bit, recorded to a 20-bit Mitsubishi X-86 2-track.
Page 47